Astra is so useful that even OpenAI can't handle it.
In the early hours of Friday, OpenAI announced the suspension of new subscriptions and upgrades to the ChatGPT Pro $200 ((Pro 20X) package for current users.
According to Thibault Sottiaux, head of the OpenAI core product and platform at Tibo, former Codex executive, the Astra model's performance is unprecedented and demand-driven.
The new supply and demand relationship is unpredictable, and OpenAI does not give a specific date for resuming sales, only indicating that capacity is increasing as quickly as possible.
This has already been confirmed by the OpenAI Help Center. I have to say that the recent temporary announcement of the big model companies has taken off, causing a bit of a stir.
Peter Steinberger, founder of Dragonfly, also said it was best to subscribe.
Astra does consume more than GPT-5.6 Sol, and OpenAI has previously reminded users that Astra will use the Work/Codex quota faster, which is about 2.5 times that of Sol. In addition, the GPT-6 Astra's most powerful Pro version model is also exclusive to Pro、Business、Enterprise users, with a Pro 5X quota of 50 items per week and a Pro 20X quota of 200 items.
On September 3, OpenAI released the GPT-6 Astra, which ranked "the strongest artificial intelligence", officially using the AGI era's claims directly, and sparked heated debate.
As for the capabilities of the new model, the users of the Station Team are the most explicit about the problem, and many in the community have already said that OpenAI has instantly snatched the users of Anthropic. OpenAI seems to be pleased with this as well. On September 9, Tibo posted a warning saying, "We're pulling all the leverage that people don't have before for Astra... if we continue this, we may have to suspend new Pro subscriptions".
Just not thinking about top-notch AI companies, but also starting a limited stream of core products, has also sparked the discontent of developers.
We know that ChatGPT has subscription ratios like Go, Plus, Pro 5X and so on, so why is OpenAI the most expensive?
It seems that in OpenAI's current payment system, the Pro 20X, although the highest total price, the least number of users, is the cheapest unit calculation, the highest usage rate package. For OpenAI, it may be the most stressful one. From the Pro 5X to the 20X, users pay more than twice as much, but get four times as much, and the Pro 20X also includes unlimited benefits such as desktop voice.
SemiAnalysis has calculated that OpenAI and Anthropic's subscriptions run to weekly limits, and then depreciate according to standard API prices. OpenAI starts to lose money on the Plus and Pro 5x with over 11.4% usage, while on the Pro 20X, the usage rate of only 5.7% enters negative gross (Anthropic is about 10%).
Furthermore, the people who actually buy the Pro 20X are the ones who actually use it. Generally, it's the long-running Agent task, the one who runs the Codex all night, the one who sends hundreds of tasks a day.
It's no wonder that people will vomit and subscribe to Coding Plan as if they were posting to themselves, just to make them work extra.
On the other hand, Pro 20X's "sell-off" also suggests that companies like OpenAI, which originally set up the Coding Plan, anticipated that "most people won't be fully satisfied", and that the actual usage behavior of heavy users after the launch of Astra has already overturned this hypothesis.
In view of the increasing use of top-notch AI, will their price estimates go up in the future?
References are given below: